Flexible Sensor Data Hub for Low-Latency SoC Workload Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing number and type of sensors in automobiles place significant burdens on System on a Chip (SoC) devices, particularly due to the high demand for processing sensor data and the need for reliable safety systems, which is exacerbated by the current reliance on DRAM for data storage and reformatting, leading to delays and bandwidth issues.
Innovation Solution
A hub is introduced to receive and distribute sensor data directly to appropriate recipients in the format they require, bypassing DRAM storage and reformatting, using a demultiplexer to filter and multiplexers to distribute data streams efficiently, reducing delays and bandwidth constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If sensor data is stored in DRAM and reformatted by each recipient, then data can be accessed by multiple processes, but processing delays increase and DRAM bandwidth is consumed
Solution Approach 1:
The hub performs preliminary actions by pre-formatting sensor data streams according to the specific requirements of each recipient before distribution. Each data stream is formatted in advance with the appropriate protocol and structure needed by its destination device, eliminating the need for recipients to perform time-consuming reformatting operations after data retrieval from DRAM.
Solution Approach 2:
The hub acts as an intermediary component between the sensor array and multiple recipient processes. It receives raw sensor data, performs centralized formatting and protocol conversion, then distributes properly formatted streams to various recipients. This intermediary role eliminates the need for each recipient to independently access and reformat data from DRAM, reducing both delay and bandwidth consumption.
2Adaptability or versatility
If sensor data is stored in DRAM, then data can be retained for multiple processes, but DRAM bandwidth issues arise
Solution Approach 1:
The hub segments the monolithic sensor data stream into multiple specialized sub-streams, each formatted for a specific recipient's requirements. Instead of having all processes compete for the same DRAM bandwidth to access complete sensor data, each process receives a dedicated segmented stream that contains only the data and formatting it needs, dramatically reducing overall DRAM bandwidth consumption.
Solution Approach 2:
The hub serves as an intermediary that performs centralized data preparation and distribution. By formatting and distributing data streams before they reach recipients, the hub eliminates repeated DRAM access cycles that would otherwise be required when multiple processes independently retrieve and reformat the same sensor data, thereby reducing DRAM bandwidth consumption while maintaining data sharing capability.
3Ease of operation
If multiple processes access sensor data independently from DRAM, then each process has data autonomy, but processing efficiency decreases
Solution Approach 1:
The hub merges the data formatting and distribution functions into a single centralized operation. Instead of multiple independent processes each performing separate data retrieval and formatting operations from DRAM, the hub combines these functions into one coordinated process that simultaneously prepares and distributes multiple formatted streams, greatly improving overall processing efficiency while maintaining process independence.
Solution Approach 2:
The hub performs preliminary formatting and preparation of data streams before distribution to multiple processes. By preparing all necessary data streams in advance with appropriate formatting, the hub enables recipient processes to immediately process the data without delay, maintaining process independence while dramatically improving overall system productivity.
Data Source
AI summary
A hub that receives sensor data streams and then distributes the data streams to the various systems that use the sensor data. Various hardware components within the hub are configured to redistribute and/or combine select data streams based on safety and/or workload issues. Any data stream determined to be faulty is not transmitted to its destination. When a receiving system is determined to have too high of a workload, select data streams may be redistributed to another system for more timely and efficient processing.