MRAM Access Control for Low-Power Sensor Node Signal Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing microcontroller units (MCUs) with low power consumption and high performance are not suitable for sensor nodes that process large numbers of signals efficiently, as they cannot effectively suppress data transfer to data centers.
Innovation Solution
A device comprising a Magnetic RAM (MRAM) with multiple memory cells, a nonvolatile CPU, a nonvolatile FPGA-ACC, and a power-gating controller, which allows for efficient data transfer and processing by using nonvolatile memory cells that do not require data backup, enabling power-gating to inactive units and parallel processing between the CPU and FPGA-ACC.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If existing low power MCUs are used, then power consumption is reduced, but processing capability for large numbers of signals deteriorates
Solution Approach 1:
The device is segmented into multiple functional units: nonvolatile CPU for control operations, nonvolatile FPGA-ACC for parallel signal processing operations, and MRAM for data storage. This segmentation allows each unit to operate independently and efficiently, with the FPGA-ACC handling computationally intensive signal processing while the CPU manages control functions, thereby achieving high signal processing capability without proportionally increasing power consumption.
Solution Approach 2:
The power-gating controller implements periodic power supply control, turning power on only when needed for specific memory cells or functional units. This periodic activation allows the system to maintain low average power consumption while providing high processing capability on demand, resolving the contradiction between continuous high performance and low power consumption.
2Loss of information
If data is transferred to data center, then information is extracted, but data transfer power consumption increases
Solution Approach 1:
The nonvolatile FPGA-ACC performs preliminary information extraction and processing operations locally within the sensor node before data transfer. By pre-processing signals and extracting useful information locally, the system reduces the volume of data that needs to be transferred to the data center, thereby reducing transfer power consumption while maintaining effective information extraction.
Solution Approach 2:
The nonvolatile FPGA-ACC acts as an intermediary processing unit between the sensor inputs and the data center. It performs intermediate signal processing and filtering operations, extracting useful information locally before transmission, which reduces the burden on data transfer and minimizes energy loss during communication.
3Reliability
If nonvolatile memory is used, then data backup requirement is eliminated, but memory access speed may deteriorate
Solution Approach 1:
The device merges nonvolatile memory (MRAM) with nonvolatile logic units (CPU and FPGA-ACC) into an integrated architecture. This integration allows the system to leverage the nonvolatile nature of MRAM for data retention while achieving high-speed access through the nonvolatile logic units that can directly interface with the memory without traditional volatile memory intermediaries, thus maintaining both reliability and speed.
Data Source
AI summary
The present invention provides an access controller, and a data transfer method. The access controller controls accesses to the MRAM by reading data in advance and backing up the data when data is to be read from the MRAM.