REGMA.
Technologies
Modern embedded device, machines and systems usually have one or several central control units. Most often these are multi-core MCUs, but sometimes high performance systems require inclusion of FPGA or GPU devices, often packed in the form of System-on-Module (SoM). This page explains main differences, and pros and cons of each device.
Example: Embedded Device System Architecture
We present a hypothetical case of medical device where complete system consists of FPGA, MCU and GPU devices. Each component performs the task at which it excels, showcasing that complex devices are usually assembled from several processing devices.
MCU
The main high-end micro-controller coordinates the whole system: it configures up FPGA, calculates parameters for analog front end, configures GPU, controls DMA to transfer stream data directly from FPGA device into memory, and evaluates user inputs. It connects into internet and manages external memory devices.
FPGA
Ultrasound probe contains large number of transducers that convert sound waves into electrical signal. Each transducer output is connected to it's own ADC channel which samples the output voltage at 200 MSamples/s. Assuming 8 bit resolution and parallel arrangement of 64 ADCs, data is generated at 12.8 GB/s. FPGA is use to interface all ADCs and process this vast amount of data. Filtered and interpreted ultrasound information is then forwarded to memory device at a much lower rate in a form of 3D coordinates and intensity.
GPU
GPU receives user settings from MCU defining what should be displayed (e.g. content, orientation, contrast, brightness,...). From FPGA it receives numerical 3D data points, and renders appropriate image to be displayed on a screen in human readable form.
Optionally, it can run a neural network to detect possible diseases.
Low-end MCU
Low-end MCU has low processing capabilities but is power savvy. In device stand-by mode it monitors start button to detect when user wants to turn on the machine. Upon turn on request, it enables power for FPGA, GPU and main MCU. During normal operation it serves as a watchdog by monitoring that every subsystem is operating as desired.