Figure 2 – Block diagram of Analog Devices MAX78000 AI MCU with CNN Accelerator
power consumption of 24uW/MHz Active Mode Current and operates from a 3V supply to maximize battery life. Analog Devices 40nm process technology devices (MAX3265x, MAX3266x and MAX3267x) all operate at 60 uW/MHz or better.
A case in point is the MAX32670, an Ultra Low Power MCU powered by an Arm Cortex M4 with Floating Point Unit is ideal for Industrial and IoT applications. Excellent energy efficiency is achieved through various operation modes designed to maximize the life of battery-operated applications. In Active mode with a 12MHz system clock, 0.9V core voltage and 160kB of memory retained, power consumption as low as 2mW is achievable on a single supply. Power consumption of 3.4uW is achievable in Deep Sleep and 0.34uW in Backup mode. All peripherals are available in the Active and Sleep modes.
Bluetooth Low Energy (BLE) MCU devices Bluetooth Low Energy (BLE) radio is integrated into many variants including the MAX32655, MAX32665-68 and MAX32680 devices. Additionally, MAX32680 is an ultra-low power Arm Cortex M4F device featuring a precision analog front-end (AFE) and a BLE 5.2 radio. This MCU integrates a RISC-V coprocessor which supervises the BLE radio taking this load off the main MCU core. Analog peripherals such as Sigma- Delta ADCs with up to 24-Bit resolution and a 12-Bit DAC are also included with the MAX32680.
Analog Devices MAX78000 and MAX78002 are AI MCUs that have been designed for AI at the Edge in battery operated applications. These MCUs implement a hardware based Ultra Low Power Convolutional Neural Network (CNN) Accelerator which is dedicated for inferencing. The CNN accelerator operates at 50MHz and can perform an AI inference in milliseconds using just microjoules of energy. The MAX78000 AI MCU with CNN Accelerator and MAX32630, a 96MHz Arm Cortex-M4F MCU, where trained to inference the MNIST dataset which is a large database of handwritten digits that is commonly used for training various image processing systems. Figure 3 shows a comparison of the inference speed and energy utilization for three scenarios. It is evident that the hardware CNN accelerator is superior to a firmware implementation. For a typical operation, the MAX78000 is at least 400 times faster and uses a factor of 1000 times less power. Realtime applications could be realised with inference times of a few milliseconds.
The CNN architecture is highly flexible, allowing networks to be trained in conventional toolsets like PyTorch and TensorFlow, then converted for execution on the MAX78000 using tools available from Analog Devices.
The internal architecture of the MAX7800 is defined by two distinct blocks whereby the MCU section is responsible for efficient movement of data from the connected sensors to the CNN Accelerator which operates independently once loaded
with data. These AI devices are equipped with interfaces that support audio and video capture such as a MIPI CSI-2 interface for video cameras on the MAX78002.
Design support
Analog Devices provide the Maxim Micros Software Development Kit (MaximSDK) for the Arm Cortex M4F and AI MCUs, these are available for Windows, Macintosh, and Linux operating systems. Software development is also supported in other development packages from Keil, IAR Systems, Eclipse and Arm MBED. For hardware development, evaluation kits and rapid prototype feather boards for all MCU variants are also available.
In addition, Anglia offers support for customer designs with free evaluation kits, demonstration boards and samples of Analog Devices products via the EZYsample service which is available to all registered Anglia Live account customers. Anglia’s engineering team are also on hand to support developers with MCU designs and can offer advice and support at component and system level. This expertise is available to assist with all aspects of product design, providing hands on support and access to additional comprehensive Analog Devices resources including technical application notes, whitepapers, reference designs and online tools to help developers select the right product for the application.
Scan the QR code or visit
www.anglia-live.com to see the full range of Analog Devices products available from Anglia.
* All brand names or trademarks used are copyright of their respective owners
Figure 3 – Inference speed and energy utilisation comparison between the MAX7800 and MAX32630
www.cieonline.co.uk Components in Electronics May 2023 11
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66
