Communication Interface for SoC Data Transmission Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In communication interfaces for Systems-on-Chip (SoCs) and Systems-in-Package (SiPs), the existing methods for transmitting data through interconnection networks, such as Network-On-Chip (NoC), face inefficiencies due to the need for additional interfaces and increased latency when handling DMA communication channels, especially when dealing with DDR memories that require row precharge operations and cache management, leading to slower access times and additional memory requirements.
Innovation Solution
A communication interface that allows for the transmission of data segments as chunks, where the interface can start transmission before all data are received if certain criteria are met, such as elapsed time exceeding a threshold or a sufficient amount of data being received, and includes features like dual FIFO memories for control and payload data, enabling chunk-based transmission with identifiers to prevent interleaving with other circuits.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data transmission waits for complete data reception before starting, then data integrity is ensured, but transmission latency increases
Solution Approach 1:
The communication interface performs preliminary actions by validating data integrity through checksum verification and setting up transmission parameters before the entire data transfer completes. This allows the transmission to start earlier while ensuring reliability through pre-established error checking mechanisms.
Solution Approach 2:
The data transmission is segmented into chunks that can be validated and prepared for transmission independently. Each chunk undergoes integrity checks and can be queued for transmission before the complete data set arrives, reducing overall latency while maintaining data integrity through segment-level validation.
2Reliability
If all data are received before transmission starts, then data completeness is guaranteed, but memory requirements increase
Solution Approach 1:
Data is divided into manageable chunks that can be processed and transmitted separately. Each chunk requires only minimal buffer memory for validation and queuing, rather than requiring the entire data set to be stored in memory simultaneously. This segmentation approach maintains data completeness through structured chunk management while dramatically reducing peak memory requirements.
Solution Approach 2:
The system performs preliminary validation and chunk preparation in advance, allowing transmission to begin before complete data reception. This preliminary processing enables the system to guarantee data completeness through validation mechanisms while avoiding the need to store all data in memory before transmission starts.
3Adaptability or versatility
If additional interfaces are added for DMA communication, then communication functionality is improved, but device complexity increases
Solution Approach 1:
The communication interface is designed with multi-functionality to handle both traditional byte-by-byte communication and DMA chunk-based transmission through a unified architecture. This universal interface approach provides enhanced communication functionality while avoiding the need for separate dedicated interfaces for different communication modes, thereby controlling device complexity.
4Productivity
If data transmission is interrupted and interleaved with other circuits, then network utilization is improved, but chunk atomicity is compromised
Solution Approach 1:
The transmission system dynamically manages chunk atomicity through configurable interruptibility parameters. Chunks can be marked as atomic or non-atomic, allowing the system to adaptively balance network utilization against atomicity requirements based on application needs. This dynamic approach enables flexible control over the trade-off between productivity and reliability.
Solution Approach 2:
Different quality levels of atomicity protection are applied to different chunks based on their specific requirements. Critical chunks receive full atomicity protection through dedicated transmission slots, while non-critical chunks allow interleaving for improved network utilization. This local quality differentiation maintains reliability where needed while maximizing overall network productivity.
Data Source
AI summary
A communication interface couples a transmission circuit with an interconnection network. The transmission circuit requests transmission of a predetermined amount of data. The communication interface receives data segments from the transmission circuit, stores the data segments in a memory, and verifies whether the memory contains the predetermined amount of data. When the memory contains the predetermined amount of data, the communication interface starts transmission of the data stored in the memory. Alternatively, when the memory contains an amount of data less than the predetermined amount of data, the communication interface determines a parameter that identifies the time that has elapsed since the transmission request or the first datum was received from the aforesaid transmission circuit, and verifies whether the time elapsed exceeds a time threshold. In the case where the time elapsed exceeds the time threshold, the communication interface starts transmission of the data stored in the memory.